Mathematisch-Naturwissenschaftliche Sektion Fachbereich Psychologie
When good intentions are not enough:
Moderators of the implementation intention effect.
Dissertationsschrift
zur Erlangung des akademischen Grades Doktor der Naturwissenschaften
Vorgelegt im Juni 2007 von
Georg Odenthal
Erstgutachter: Prof. Dr. Peter M. Gollwitzer Zweitgutachterin: Prof. Dr. Sabine Sonnentag
Termin der mündlichen Doktorprüfung: 18. Januar 2008
It is not good to settle into a set of opinions. It is a mistake to put forth effort and obtain some understanding and then stop at that. At first putting forth great effort to be sure that you have grasped the basics, then practicing so that they may come to fruition is something that will never stop for your whole lifetime. Do not rely on following the degree of understanding that you have discovered, but simply think, “This is not enough.” One should search throughout his whole life how best to follow the Way. And he should study, setting his mind to work without putting things off. Within this is the Way.
Hagakure - The Way of the Samurai
An dieser Stelle möchte ich mich bei all den Personen bedanken, die mir beim Erstellen dieser Arbeit behilflich waren.
Dank gilt zunächst einmal meinem Doktorvater Prof. Dr. Peter M. Gollwitzer dafür, dass er mir die Möglichkeit gegeben hat, die Dissertation an seinem Lehrstuhl zu schreiben. Frau Prof. Dr. Sabine Sonnentag danke ich herzlich für die Übernahme der Zweitbegutachtung dieser Arbeit.
Darüber hinaus möchte ich mich ebenfalls bei Dr. Frank Wieber für seine Hilfe bei den Analysen und seine Tipps zur Ausarbeitung bedanken.
Dank gebührt auch meinen Kollegen in Konstanz für die nette Gesellschaft und die soziale Unterstützung bei der Arbeit und meinen beiden Hiwis Michi und Verena H.
für ihre Hilfe bei der Datenerhebung.
Den Teilnehmern der Kolloquien des Lehrstuhls für Sozialpsychologie und Motivation an der Universität Konstanz und der Forschungsstelle für Motivationspsychologie an der Universität Hamburg bin ich dankbar für ihre Tipps und Anregungen.
Ein ganz besonderer Dank geht an meine Eltern für ihre jahrelange finanzielle und moralische Unterstützung.
Meiner Partnerin Verena bin ich unendlich dankbar für ihre Unterstützung und Geduld über all die Jahre und für so vieles mehr!
ZUSAMMENFASSUNG _______________________________________________ 1 ABSTRACT ________________________________________________________ 3 LIST OF TABLES AND FIGURES _______________________________________ 9 INTRODUCTION ____________________________________________________ 4 STUDY 1: SIMPLE VERSUS COMPLEX IMPLEMENTATION INTENTIONS _____ 10 OVERVIEW _______________________________________________________ 10 METHOD _________________________________________________________ 13 RESULTS ________________________________________________________ 16 DISCUSSION______________________________________________________ 23 CONCLUSION _____________________________________________________ 26 STUDY 2: EFFECT OF POSITIVE AND NEGATIVE INCENTIVE ON
IMPLEMENTATION INTENTIONS _____________________________________ 27 OVERVIEW _______________________________________________________ 27 METHOD _________________________________________________________ 29 RESULTS ________________________________________________________ 31 DISCUSSION______________________________________________________ 37 CONCLUSION _____________________________________________________ 43 STUDY 3: INFLUENCE OF GOAL-COMMITMENT ON THE IMPLEMENTATION INTENTION EFFECT________________________________________________ 44 OVERVIEW _______________________________________________________ 44 METHOD _________________________________________________________ 47 RESULTS ________________________________________________________ 50 DISCUSSION______________________________________________________ 56 CONCLUSION _____________________________________________________ 59 GENERAL DISCUSSION_____________________________________________ 60 REFERENCES ____________________________________________________ 66 APPENDIX A ______________________________________________________ 73 APPENDIX B _____________________________________________________ 101 APPENDIX C _____________________________________________________ 123
List of Tables and Figures
Tables
Table 1___________________________________________________________ 17 Table 2___________________________________________________________ 18 Table 3___________________________________________________________ 20 Table 4___________________________________________________________ 22 Table 5___________________________________________________________ 32 Table 6___________________________________________________________ 36 Table 7___________________________________________________________ 51 Table 8___________________________________________________________ 52 Table 9___________________________________________________________ 55
Figures
Figure 1 __________________________________________________________ 15 Figure 2 __________________________________________________________ 19 Figure 3 __________________________________________________________ 34 Figure 4 __________________________________________________________ 35 Figure 5 __________________________________________________________ 37 Figure 6 __________________________________________________________ 54
Zusammenfassung
Es wurde bereits hinreichend nachgewiesen, dass das Verbinden von Zielen mit Vorsätzen (‚Wenn Situation X auftritt, dann will ich Verhalten Y ausführen’) eine wirksame selbstregulatorische Strategie zur Verbesserung der Zielrealisierung darstellt. Aktuelle Forschungsergebnisse deuten darauf hin, dass die förderliche Wirkung von Vorsätzen auf die Zielrealisierung durch situationale und individuelle Faktoren moderiert werden kann. Diese Befunde wurden jedoch entweder
methodologisch ungenau oder in Feldstudien erhoben, bei denen keine vollständige Kontrolle über Störvariablen herrscht.
In der vorliegenden Arbeit wurde der Einfluss von drei möglichen Moderatorvariablen auf den Vorsatzeffekt in einem kontrollierten Laborumfeld untersucht. In der ersten Studie wurde die Komplexität der im Vorsatz spezifizierten Handlung variiert, d.h. die Teilnehmer fassten sich entweder einfache oder komplexe Vorsätze und
bearbeiteten anschließend dieselbe Aufgabe. Das Ergebnis dieser Studie war, dass die Art und Weise, wie ein Vorsatz formuliert wird einen Einfluss auf dessen
Wirksamkeit hat. Teilnehmer, die sich einen komplexen Vorsatz vornahmen lösten weniger Aufgaben richtig als die Teilnehmer, welche sich einen einfachen Vorsatz setzten. Dieser Einfluss war jedoch nur auf männliche Teilnehmer der Studie beschränkt, bei Frauen wurden keine Veränderung der Leistungen gefunden, was vermutlich an der Art der Aufgabe lag. In Studie 2 wurde der Handlungsanreiz für die zielgerichtete Handlung manipuliert, um zu untersuchen, ob Vorsätze bei geringem Handlungsanreiz nicht mehr ihre Leistungsförderliche Wirkung ausüben können.
Diese Annahme konnte jedoch nicht bestätigt werden. Stattdessen deuteten die Befunde an, dass ein geringer Handlungsanreiz unabhängig der Vorsatz-Bedingung zu einer allgemein erhöhten Reaktionsgeschwindigkeit führte. Darüber hinaus verursachte eine unerwartete Störvariable Komplikationen bei der Auswertung, aufgrund derer die Ergebnisse dieser Studie nicht verallgemeinert werden können.
Mögliche Ursachen für die unerwarteten Befunde und den Einfluss der Störvariable werden diskutiert und Empfehlungen für die weitere Forschung mit diesem
Paradigma werden besprochen. In der dritten Studie wurde das Verpflichtungsgefühl der Teilnehmer gegenüber ihrem Ziel experimentell manipuliert. Es wurde erwartet, dass Vorsätze nur bei einer hohen Zielverpflichtung eine Leistungssteigernde
Wirkung ausüben können. Diese Annahme konnte durch die Daten zum Teil bestätigt
werden. Nur die Teilnehmer, die eine hohe Verpflichtung gegenüber ihrem Ziel verspürten konnten bei schweren Aufgaben von dem gefassten Vorsatz profitieren und eine bessere Leistung erzielen. Bei einfachen und mittelschweren Aufgaben zeigten sich keine Leistungsunterschiede zwischen den Teilnehmern der
verschiedenen Vorsatzgruppen. Die Resultate dieser drei Studien liefern wichtige Anhaltspunkte für die Moderation des Vorsatzeffekts auf die Zielrealisierung durch situationale und individuelle Faktoren. Die Bedeutung dieser Ergebnisse für Labor- und Feldstudien zum Vorsatzeffekt wird diskutiert und Richtlinien für die zukünftige Forschung werden präsentiert.
Abstract
Furnishing ones goals with implementation intentions (‚If I encounter situation X, then I’ll perform behavior Y’) has been shown to be an effective self-regulatory strategy to improve goal-achievement. Recent research has suggested that the beneficial effect of implementation intentions on goal-success can be moderated by situational and individual factors. But these findings had either methodological inaccuracies or were conducted in the field without the ability to control for confounding factors.
The present research explored the impact of three possible moderators on the implementation intention effect in a controlled laboratory setting. In Study 1 it was found that the way in which an implementation plan was phrased moderated its effectiveness. These results were limited to male participants though, the performance of female participants was not moderated by the phrasing of the implementation intentions. Study 2 found some support for the influence of an incentive to act on the effectiveness of implementation intentions. But the results pointed into the opposite direction to what was hypothesized. Possible reasons for this unexpected finding are discussed. Finally, in study 3 the goal-commitment of the participants was experimentally manipulated. It was found that high goal-commitment is a necessary prerequisite for the effectiveness of implementation intentions for difficult tasks. Overall these findings provide support for the moderation of the implementation intention effect on goal achievement by situational and individual factors. The importance of these findings for laboratory and field studies using implementation intentions is discussed and directions for future research are presented.
Introduction
People often do not reach the goals they are striving for. Even when holding strong attitudes towards performing a specific behavior, in many cases these actions are not carried out. For example in a classical study, LaPiere (1934) found that over 90 percent of hotel and restaurant owners responded that they would not serve Chinese customers, but when LaPiere visited 251 hotels together with a Chinese couple only one actually refused to accommodate them. This discrepancy between a person’s intention to act and the actual behavior is referred to as the intention-behavior gap (cf. Gollwitzer & Sheeran, 2005).
Intentions are defined as “instructions people give to themselves to behave in certain ways” (Triandis, 1977, p. 203) and usually have the form “I intend to do/ to achieve X”. Forming an intention marks the end point of a decision making process, because it signals that the individual has settled on a course of action and has discarded all other alternatives.
A person’s intention to act is the most immediate and important predictor of subsequent action as posited by theories like the theory of planned behavior (TPB, Ajzen, 1991), the protection motivation theory (PMT, Rogers, 1975, 1983) and Triandis’ (1980) attitude-behavior theory. Indeed, several findings show that intentions do predict behavior at least to some degree (Ajzen & Fishbein, 1977;
Armitage & Conner, 2001; Milne, Sheeran, & Orbell, 2000). In order to quantify the relationship between intentions and action Sheeran (2002) carried out a meta- analysis of meta-analyses on the intention-behavior relationship and found that on average intentions account for 28% of the variance in behavior. According to Cohen’s (1992) power primer an R2 = .28 constitutes a “large” effect size, which suggests that intentions are “good” predictors of behavior (Sheeran, Milne, Webb, & Gollwitzer, 2005, p. 277). But that still leaves a lot of variance unaccounted for.
A number of ways to improve goal-attainment have been proposed. First, there are approaches that try to frame intentions or goals differently in order to enhance goal success. For example, Ajzen and Fishbein (1980) found that several factors like attitudes towards a behavior, subjective norms and perceived behavioral control have an influence over the strength of an intention to act and account for considerable variance in the actual behavior (Ajzen, 1991). Dweck (1996)
demonstrated that when individuals frame their goals as learning goals in contrast to
performance goals the attainment of the goal was more likely. Furthermore goal attainment can also be improved if people set themselves challenging and specific goals instead of mere “do your best” goals (Locke & Latham, 2002). Higgins (1997) found that motivation and goal commitment are higher when people use goal pursuit means that fit their regulatory orientation of either attaining a desired outcome or avoiding an undesired result. The dividing of distal goals into proximal subgoals has also been shown to be a reliable method of improving performance and developing a stronger sense of self-efficacy and an improved intrinsic motivation (Bandura &
Schunk, 1981). Furthermore Ryan and Deci (2000) found that goals based on the three psychological needs of competence, autonomy and relatedness are more often achieved than goals based on other needs.
Second, a different line of research is concerned with self-regulatory skills that are necessary for successful goal-attainment. Self-regulatory skills are important in initiating goal-directed behaviors, in shielding an ongoing goal-striving from
distractions, in coping with competing goals and in trying to prevent falling back into bad habits (Gollwitzer, 1999). Self-regulation exceeds the merely motivational issues of planning and setting of goals by adding a volitional (willful) component to the goal- striving process. The distinction between motivational and volitional components of goal striving is described by the model of action phases (Gollwitzer, 1990;
Heckhausen, 1989) which postulates four phases – two motivational, two volitional – that are passed through during goal attainment. The first phase, which is called
“predecisional” phase (Gollwitzer, 1996, p. 289), is volitional and consists of deliberating the feasibility and desirability of given wishes and the setting of preferences. The end of this phase marks the setting of a goal to pursue. In the second phase, the “preactional” phase, concrete steps on how one wants to reach the selected goal are planned, that is the “when”, “where” and “how” of the goal- directed behavior are determined. The end of this phase marks the initiation of actions to reach the goal. During the third phase, the “actional” phase, the planned goal-directed behaviors are carried out until goal completion. Both of these phases are motivational, while the last phase is again volitional. In this last phase, the
“postactional” phase, the actor evaluates whether the goal has been fulfilled or whether more steps to goal-completion are necessary.
The self-regulatory strategy of forming implementation intentions (Gollwitzer, 1999) is based on the model of action phases. Implementation intentions are
subordinate to goal intentions (“I want to achieve Z”) and specify the “when”, “where”
and “how” of goal-directed actions. That is, implementation intentions are essentially the result of the preactional phase of the model of action phases and have the form
“If situation X arises, then I will perform behavior Y”. Implementation intentions link the goal-directed behavior to an anticipated situational context, which is assumed to lead to the automatization of the intended goal-directed behavior once this situational context is encountered. Action initiation becomes swift, efficient and does not require conscious intent (Gollwitzer, 1999, p. 495). This automatization is rooted in the hightened accessibility of the situational cue specified in the if-part of the
implementation intention. Aarts, Dijksterhuis and Midden (1999) found evidence to support this assumption by demonstrating that the effects of planning on goal completion are mediated by a hightened accessibility of environmental cues related to a goal-completion task. Over and above that Webb and Sheeran (2004,
Experiment 1) demonstrated that forming implementation intentions leads to
improved cue detection compared to forming mere goal intentions or to familiarizing oneself with the situational cue.
Over the past 15 years a great number of studies have demonstrated the effectiveness of implementation intentions in warding off self-regulatory problems like getting started, shielding ongoing goal-pursuits from unwanted influences or falling back into bad habits (for an overview see Gollwitzer & Sheeran, 2005). For example, implementation intentions have been shown to improve attendance for cervical cancer screening (Sheeran & Orbell, 2000), use of vitamin supplements or
performance of breast self-examinations (Orbell & Sheeran, 2002; Prestwich et al., 2005), fulfillment of home assignments (Gollwitzer & Brandstaetter, 1997, experiment 2), eating healthier food (Armitage, 2004; Nooijer, de Vet, Brug, & de Vries, 2006), performing physical activity (Latimer, Ginis, & Arbour, 2006; Luszczynska, 2006;
Walsh, da Fonseca, & Banta, 2005) or enhance use of public transportation (Bamberg, 2000) and promote recycling behavior (Holland, Aarts, & Langendam, 2006). Gollwitzer and Sheeran (2005) recently analyzed 94 independent tests of the effect of implementation intentions on the rate of goal attainment using the method of meta-analysis and found a medium-to-large (d = .65) effect size, indicating that implementation intentions are indeed a powerful self-regulatory strategy to promote goal achievement.
So far, implementation intentions have been proven to be effective in many areas of life by laboratory as well as field experiments. However, some studies have recently challenged the universality of the implementation intention effect on
successful goal achievement. For example Koestner and his colleagues (Koestner et al., 2006; Koestner, Lekes, Powers, & Chicoine, 2002; Powers, Koestner, & Topicu, 2005) found that implementation intention interventions provided only short-term effects, but did not improve goal-achievement over longer periods of time. But this research is in contrast to studies that did find long-term effects of implementation intentions (Holland, Aarts, & Langendam, 2006; Luszczynska, 2006; Orbell &
Sheeran, 2002), indicating that more research is needed to provide conclusive evidence for the durability of the implementation intention effect.
Moreover, several moderators of the implementation intention effect have been suggested. For example, Sheeran, Webb and Gollwitzer (2005, study 1) found that goal-strength moderates the effect of implementation intentions on goal-
attainment, meaning that only participants with strong goals were able to profit from forming implementation intentions. Koestner, Lekes, Powers and Chicoine (2002) discovered that implementation intentions improved goal-attainment only if the goals were self-concordant, that is if the goal reflected personal interests and values and was not determined by external or internal pressures. Additionally, it was found that implementation intentions only affect performance when the task-relevant goal has been activated (Sheeran, Webb, & Gollwitzer, 2005, study 2) and that the strength of the commitment to an implementation intention also moderates its effectiveness (de Nooijer, de Vet, Brug, & de Vries, 2006).
Over and above that, three experiments reported by Dewitte, Verguts and Lens (2003) indicate, that implementation intentions do not help to improve goal- achievement when the individual is unsure about which actions are required to reach a difficult goal. Furthermore, two studies report diminished implementation intention effects in demanding tasks like high time-pressure tasks (Betsch, Haberstroh, Molter,
& Glockner, 2004) and high cognitive load (Einstein, McDaniel, Williford, Pagan, &
Dismukes, 2003). But in the latter study, participants were only given very little time to form their implementation intentions and were also not explicitly told to create their plans in the typical if-then format. Thus, methodological problems could also account for the absence of an implementation intention effect in the Einstein et al. (2003) study. Since a number of other studies did find implementation intention effects under
cognitive load (Brandstaetter, Lengfelder, & Gollwitzer, 2001; Lengfelder &
Gollwitzer, 2001) or time-pressure (Henderson, Gollwitzer, & Oettingen, 2007) it is possible that different variables than the task-demand impaired the effectiveness of implementation intentions on goal-achievement in the studies above.
Individual differences may also account for some variation in the strength of the implementation intention effect, as Gollwitzer and Sheeran (2005, pp. 106-107) have already suggested. Whereas there are two studies that found beneficial interactions of implementation intentions and personality variables (Brandstaetter, Lengfelder, & Gollwitzer, 2001; Lengfelder & Gollwitzer, 2001) there are another two studies that found a negative interaction. First, according to research by Powers, Koestner and Topicu (2005) socially prescribed perfectionists, that is people who have the need to attain standards or expectations prescribed by significant others, are not able to profit from furnishing their goals with implementation intentions. In contrast, the data of Powers and his collaborators suggests that forming
implementation intentions even resulted in a significant backfire on the goal progress for socially prescribed perfectionists. The authors hypothesize that forming
implementation intentions may initiate a process of self-criticism and hypervigilance for potential critical judgment (Powers, Koestner, & Topicu, 2005, p. 909) which interferes with the self-regulatory function of implementation intentions.
Secondly, Downie, Koestner, Horberg and Haga (2006) provide some
evidence that implementation intentions interact with the self-construal - the way one conceptualizes oneself in relation to others (Markus & Kitayama, 1991) - indicating that implementation intentions have stronger effects when one has an independent self-construal compared to interdependent self-construals. This study points to culture as a possible moderator of the implementation intention effect and suggests, that westerners who usually have independent self-construal can profit more from forming implementation intentions than easterners who rather have interdependent self-contruals.
Though not all of the studies cited above provide conclusive results on the influence of additional variables or situational constraints on the implementation intention effect, evidence for such moderators is accumulating. Most of these studies are field studies, where the researchers have little control over environmental and other influences on the behavior of participants. In some experiments (Downie, Koestner, Horberg, & Haga, 2006; Koestner et al., 2006; Powers, Koestner, &
Topicu, 2005) the participants had to form two or three implementation intentions for up to four different goal. But in the analysis the effects of the implementation
intentions were pooled so that the individual contribution of each implementation intention could not be accounted for. Also the participants created the implementation intentions themselves, but the studies do not report whether the feasibility of these plans and instrumentality of the behavior for the goal was verified. Thus, the failure to find an implementation intention effect could have also just been the result of a
memory overload or an insufficient quality of the implementation plans.
These findings warrant additional research in a more controlled environment in order to double-check the results and to rule out confounding factors like the
suggested memory effects or plan quality. The following three studies are aimed to resolve some of the issues from the research cited above and to provide additional insight into the functionality of the implementation intention effect. The quality of the formulation of the implementation intention is examined in first experiment. Study 2 looks at the impact of a decreased incentive to act on the performance in a typical implementation intention task. Finally, in study 3 the influence of goal-commitment on the efficiency of implementation intentions is investigated.
Study 1: Simple versus complex implementation intentions
Overview
The first study deals with the way in which action plans are converted into
implementation intentions. That is, how the action is put into words and whether the specific formulation is helpful for carrying out the planned action or interfering with the execution of that action.
Implementation intentions consist of two parts, the if-part in which attributes of a situation in which a specific goal directed action is to be performed are specified and the then-part in which the action itself is defined. Improper formulation can impede either part of the implementation intention. On the one hand good situations to act out a goal-directed behavior can be missed if the if-part of an implementation intention is inappropriately formulated and does not clearly define the critical cues of this situation or is not even considering this specific situation as a good point in time to act. On the other hand the action given in the then-part may not be carried out, because the specified action is ambiguous, too complicated to perform (i.e. when the action overstrains a person’s self-regulatory abilities), too complex (i.e. when the action involves too many interdependent steps) or is just not helpful for the respective goal intention or situation. For example it was found in studies with adults (Schaal &
Gollwitzer, 2000) and children (Gawrilow, 2004) that when individuals are distracted while solving math problems, forming distraction-inhibiting implementation intentions lead to better results than using task-facilitating ones. Thus, just furnishing one’s goal with an implementation intention, may not lead to improved performance. The
content and the formulation of the implementation intention also matters. If the if- and the then-parts of an implementation intention are not instrumental for the goal or put into too many words to be fully remembered, the proposed automatization of goal- directed behavior (Gollwitzer, 1999) cannot be established. The focus of the present study is to analyze whether the effect of an implementation intention is impaired if the specified action in the then-part is too complex to be beneficial for the goal.
The use of the term complex if often confused with the use of the term complicated. Even Merriam-Webster’s Collegiate Dictionary (Merriam-Webster's Collegiate Dictionary, 2003) and the Oxford English Dictionary (Simpson & Weiner, 1991) use the terms complex and complicated synonymous with one another. Both
terms are derived from Latin, complex comes from complexus which means embrace or grasp and complicated comes from complicare which means to fold together.
Thus, there is a small but important difference between those two terms. A complicated structure is something that is folded together and of which not all
aspects are visible while a complex structure is composed of two or more interrelated parts that are dependent on one another and whose combination is more than just the sum of its parts (Anderson, 1972). In short, complicated is the opposite of simple whereas complex is the opposite of independent.
In cognitive psychology and neuropsychology stimulus complexity is often an important factor when studying human or animal perception (Minda & Smith, 2001;
Sary et al., 2004). There are several definitions of complexity: Classic definitions of the term complex refer to the number and randomness of the units making up a picture (Hogan, 1975, p. 33). A more recent publication defines complexity of a visual stimulus as the number of dimensions used in this stimulus, where dimensions refers to the number of wavelengths, tones of different frequencies and the spatial distance of elements in an image (Fetterman, 1996). In the behavioral sciences a behavior is called complex the more neural structures in the brain are needed to carry it out.
Thus, behaviors such as feeding, learning, language and culture are defined as being complex (Greenberg, Partridge, Weiss, & Haraway, 1999). In general stimuli are referred to as complex, when they are composed of multiple interrelated units that are not easy to process.
The strategies used in this study were designed to combine these
requirements in order to create implementation intentions with varying complexity. It was important that the stimuli for which these strategies are intended would be the same or at least very similar to one another to rule out cognitive explanations for the expected performance differences between subject groups. That is the stimuli should take up the same visual area, do not differ on contrast, color or form but still provide enough differences so that different behaviors are needed to solve tasks using these stimuli. Therefore, three different types of the same stimulus had to be used to match the strategies, so that each strategy would fit one of the three stimulus types, but would still be helpful for the other two. Mental rotation (Shepard & Metzler, 1971) offered a possibility to satisfy these requirements. The figure is always the same, but by varying the rotation axis the behaviors needed to match the stimulus with the comparison figure would be different. Two stimulus types and the according
strategies were set for the rotation around a single axis (the x-axis and the y-axis respectively) and the third type of stimulus was rotated around both of these axes. By adding a stimulus that was rotated around two axes the behavior could be made more complex, since two steps have to be performed to match the stimulus with the comparison figure. Since rotating the stimulus around one axis will not suffice to match the stimulus with the comparison stimulus both rotations are necessary and dependent on one another. Over and above that the participant has to decide which rotation axis to try out first, which means that this strategy is not just a combination of the other two, but requires additional cognitive resources to make this decision. So the third strategy is not just a combination of the first two strategies, but also includes the request to decide which axis to try first. But deciding which rotation axis to use first takes time and that should be reflected in the response time of participants using the complex strategy. Moreover when the subjects have to visualize two successive rotations in their mind’s eye, it is required to keep the result of the first rotation in mind while performing the second. This puts more cognitive load (Sweller, 1994) on the participants since they have to retain the result of the first rotation in their minds while performing the second rotation. Therefore, using the strategy with two rotation axes should lead to more errors due to the additional load than using strategies with only one rotation axis.
Thus, it is expected that only participants who furnish their goal, to solve a lot of tasks, with a simple implementation intention will find more matching shapes than participants who furnish their goal with the complex implementation intention or who just set themselves the goal alone.
When the strategy in the implementation intention matches with the rotation axis of a stimulus, then the above predicted effect should be especially strong. For example, if subjects who use the strategy to rotate the stimulus around the x-axis work on a task in which the shapes are rotated around the x-axis then they should find more matching shapes than subjects using one of the other two implementation intentions or just the goal intention.
Shepard and Metzler (1971) found that the average response time to the mental rotation tasks increased linearly with the angular difference between the two shapes, that is the subjects found it more difficult to solve tasks with greater angular difference between the two shapes than with smaller differences. According to
Gollwitzer and Sheeran (2005, p. 101) engaging in if-then plans should be especially
advantageous when task features make it difficult to execute the behavior. Therefore, it is also hypothesized that the anticipated implementation intention effect in this study should become stronger with increasing task difficulty.
Method
Participants and design. 124 students from the University of Konstanz participated in this study in exchange for 5 Euros (approximately 4 US$) or course credit (only Psychology students). The sex of the participants was almost evenly distributed (65 female and 59 male) and the mean age was 23 years.
The data of two subjects was excluded from the analysis since both reported feeling sick during the study. One participant was excluded for not following the instructions and another three participants indicated that they were information science students and were familiar with the kind of wire-frame images used in the experiment. Some studies have shown that using computers and playing video games can increase spatial rotation abilities (Dorval & Pepin, 1986; Okagaki &
Frensch, 1994; Smith, 1999). So in order to avoid that the advantage of these information science students over the rest of the population would negatively
influence the analyses, the data of these three participants was also removed. Thus a total of 118 participants were subjected to a three-factorial design with the between factor goal-condition (goal, x-axis implementation intention, y-axis implementation intention vs. complex implementation intention) x the within factors task-difficulty (easy, average vs hard) and rotation axis (x-axis, y-axis, x- and y-axes).
Materials. For this study a redrawn version of one of the three-dimensional shapes from the Shepard and Metzler (1971) mental rotation test (MRT) was used.
This shape was recreated as a vector graphics object on a desktop computer and consisted of ten cubes that were connected on one or two sides with other cubes to form a shape that looked roughly like two uppercase “L” characters that are
connected to each other on one side (see Figure). The computer calculated three sets with 36 images of this shape rotated either around the x- axis, the y-axis or around both axes in 10 degree steps from 0 degrees to 350 degrees. The resulting 108 images were mirrored at the y-axis to create the non-corresponding stimuli. In a pre-study the original and the mirrored stimuli were randomly paired with one another and 19 students (ten female and nine male) had to decide whether the two presented
stimuli did or did not match. This was done in order to rate the difficulty of the shapes and exclude those shapes that were too difficult to be correctly identified. Thus 29 original and 29 mirrored stimuli were excluded from the set resulting in a new set of 158 original and mirrored stimuli. With these stimuli 108 tasks with four shapes and one criterion figure were created. The shapes were combined in a manner that resulted in three difficulty levels. Similarly to Shepard and Metzler’s study (1971) difficulty was defined as the angular difference between the four shapes and the criterion figure, that is the angular difference of all four shapes with the criterion figure was summed up and then divided by four. An average difference of 0 to 60 degrees was defined as easy, 61 to 120 degrees as average and 121 to 180 as difficult.
Procedure. The participants worked on this task on computers in separate cubicles. All instructions were given on the computer screen and the participants were allowed to call the male experimenter and to ask questions until the start of the main block. On the instruction screens the participants were informed that the
following task was a mental rotation test and an example of this task was given. After that the participants were asked to practice the basic mental rotation task in the first practice block. In this practice block 12 tasks with two shapes were presented and the subjects had to indicate whether the shapes are the same or not. After finishing this basic practice block, the participants were told that the task in the main
experiment was more difficult, because two out of four shapes had to be matched to a criterion figure. This procedure is similar to the mental rotation test used in a study by Vandenberg and Kuse (1978). It is thought of to be more difficult than the original test by Shephard and Metzler (1971) and less susceptible to ceiling effects. On the instruction screens an example of this task was presented and the subjects were told that they should try to finish each task within 15 seconds. Again the participants were allowed to practice ten tasks of this kind. Accuracy and timing feedback was provided in this second practice block. Once the practice session was over the participants were informed again that no accuracy feedback would be given during the main task.
They were also told to try to adhere to the 15 second time limit per task even though no means to double check the time was given.
Then all participants were asked to set themselves the goal to find a lot of matching shapes (i.e. “I want to find a lot of matching shapes”). Three out of four participants were randomly selected to furnish their goal with an additional
implementation intention. Two of these implementation intentions specified a simple
Figure 1
Example for a mental rotation task from the study. Shapes 1 and 3 are matching the criterion figure on top of the screen.
plan to help finding the matching shapes by either trying to mentally rotate the stimuli around the x-axis (ImpX) or around the y-axis (ImpY) every time a new task starts (e.g. “If a new task starts, then I’ll mentally rotate the stimuli around the x-axis.”). The third implementation intention (ImpC) included the plans from the former two in form of a complex new plan to select a rotation axis every time a new task starts (i.e. “If a new task starts, then I first decide whether I’ll mentally rotate the stimuli around the x- or y-axis.”). All participants were asked to rehearse their goals and plans and had to write them down on a sheet of paper that was placed next to the computer screen.
The main block consisted of 108 trials for which neither accuracy nor timing feedback was given. After the end of the main block the participants were asked a few
questions to the way they solved they tasks, whether they enjoyed the task or not and how well they thought they did on the task. Answers were given on a 7-point Likert scale. Additionally all participants had to recall their goal and implementation
intentions by writing them down on a piece of paper. Finally they had to answer to a few demographic questions.
Debriefing. Once the experiment was over, the participants were asked to get the experimenter, who then interviewed the participants about their thoughts on the experiment. Most participants expressed the thought that the experiment had
something to do with gender differences, but none were aware of the implementation intention manipulation or its expected effects on their behavior. After the interview, the experimenter fully debriefed the subjects, paid them their remuneration and thanked them for participating.
Results
Performance. Men found significantly more matching shapes during the baseline block than women, t(116) = 4.471, p < .001 (M = 74%, SD = 15 vs. M = 62%, SD = 13 respectively). Also during the main block, after the setting of goals and
implementation intentions, men found more matching shapes than women, F(1, 118)
= 9.508, p = .003 (see Table 1). The interaction of goal-condition and sex was significant as well, F(3, 118) = 3.545, p = .017, indicating that furnishing ones goal with an implementation intention leads to an improved performance. Further analysis revealed that this assumption only holds true for men who demonstrate a strong implementation intention effect, F(3, 52) = 4.450, p = .008, whereas women did not profit from furnishing their goal with implementation intentions. Contrast analyses for male participants revealed that the source of the effect were indeed the simple
implementation intentions. Participants in the simple implementation intention groups performed significantly better than participants in the goal group (weights -2 1 1 0), t(49) = 3.517, p = .001, as well as participants in the complex implementation intention group (weights 0 1 1 -2), t(49) = 2.163, p = .035.
Task-difficulty. The percentage of correctly found shapes was entered into a repeated-measures ANOVA with the between factors goal-condition (goal, x-axis implementation intention, y-axis implementation intention and complex
implementation intention) and sex and the within factor task-difficulty (easy, average and hard). Only the main effect for task-difficulty was significant, F(1.84, 202.365) = 86.474, p < .001, η2 = .440 using the Greenhouse-Geiser correction for
heterogeneous variances (Greenhouse & Geisser, 1959). A highly significant linear
Table 1
Performance of both sexes in the main block in percent of correctly matched shapes.
Goal-condition Sex M
Goal Male 69.46 (10.62)
Female 73.51 (12.75)
ImpX Male 82.95 (10.25)
Female 68.37 (14.29)
ImpY Male 79.43 (11.01)
Female 68.76 (11.01)
ImpC Male 74.13 (7.44)
Female 69.22 (11.67)
Note: Standard deviations are in parentheses.
trend for task-difficulty confirmed that the participants found fewer matching shapes the harder the task became, F(1, 110) = 137.038, p < .001, η2 = .555 (see Table 2).
Rotation Axis. To assess whether the three different implementation intentions led to an increased performance on trials where the rotation axis of the shapes
matched the strategy in the implementation intention, a repeated measures ANOVA with the between factor goal-condition and the within factor performance for each rotation axis (x-axis, y-axis and both axes) was computed for male participants. Only the main effect for rotation axis was significant, F(2, 48) = 78.089, p < .001, η2 = .765. Examination of the performance means suggests that participants found the most correct matches when the shapes were rotated around the y-axis, second most when the shapes were rotated around the x-axis and the least number of correct matches when the shapes were rotated around both axes (see Figure 2). Polynomial contrasts indicated, in support of this, a significant linear trend for rotation axis, F(1, 49) = 155.488, p < .001, η2 = .760.
Table 2
Performance of both sexes in the main block in percent of correctly matched shapes for each of the three difficulty levels.
Task-difficulty Goal-
condition Sex M
Easy Goal Male 73.55 (12.21)
Female 76.36 (13.34)
ImpX Male 86.46 (9.96)
Female 72.19 (15.30)
ImpY Male 83.10 (10.42)
Female 73.72 (11.42)
ImpC Male 76.97 (8.61)
Female 72.15 (12.87)
Average Goal Male 70.75 (10.54)
Female 75.03 (12.49)
ImpX Male 84.47 (10.09)
Female 70.19 (16.09)
ImpY Male 79.04 (11.25)
Female 69.83 (10.30)
ImpC Male 76.19 (8.12)
Female 71.43 (12.21)
Hard Goal Male 64.00 (12.45)
Female 69.03 (15.32)
ImpX Male 77.81 (12.22)
Female 62.63 (13.67)
ImpY Male 76.09 (12.63)
Female 62.62 (12.77)
ImpC Male 69.13 (8.11)
Female 63.98 (12.32)
Note: Standard deviations are in parentheses.
Contrary to the prediction, the interaction of rotation type and goal-condition was not significant. That means when the rotation axis of the shapes matched the strategy in the implementation intention the performance of the participants did not significantly improve. The means plot (Figure 2) also revealed that participants with an x-axis implementation intention displayed an overall superior performance to the
participants using the other strategies in spite of the random distribution of the subjects to the different goal-conditions. The performance means of men in the baseline block also indicate that the participants in the x-axis condition were better at this task than participants in the other three conditions (performance means in the baseline block: MGoal = 51, SD = 21; MImpX = 69, SD = 25; MImpY = 57, SD = 23 and MImpC = 49, SD = 22). But when entering the performance in the baseline block into an ANOVA with goal-condition as factor, the result was not significant (p = .169). Also when running the repeated measures ANOVA for rotation specific performance in the main block, again with the performance in the baseline block as covariate, the result was similar albeit weaker as before. The interaction of the covariate with the rotation type was far from significance with p = .923. The main effect of rotation type was
Figure 2
Average performance of male participants for the three different rotation types in each goal-condition during the main block.
again significant, F(2, 47) = 11.281, p < .001, η2 = .324 as was the linear trend, F(1, 48) = 22.789, p < .001, η2 = .322, while the interaction of rotation type and goal- condition still was non-significant.
Response times. Response time was assessed for each click on one of the four shapes. Only response times for correct responses were used to compute an index of time taken for each correct response. The index for the baseline block was subjected to a t-test with sex as the independent variable, revealing a significant reaction time difference between men and women, t(116) = 2.916, p = .004 (M = 8.53 seconds per solution, SD = 2.74 and M = 10.30 seconds per solution, SD = 3.68 respectively). Next the index for the main block was subjected to an analysis of variance with the between factors goal-condition (goal, x-axis implementation
intention, y-axis implementation intention and complex implementation intention) and sex, which also resulted in a significant main effect for sex, F(1, 117) = 5.636, p = .019. Women needed more time to select a correct shape than men (see Table 3).
Finally, the time index of the main block was computed for each difficulty level separately and then entered into a repeated measures ANOVA with the between factors goal-condition and sex and the within factor task-difficulty (easy, average and
Table 3
Average time taken (in seconds per solution) to select a correct shape of men and women in each goal-condition during the main block.
Goal-condition Sex M
Goal Male 6.28 (1.32)
Female 7.12 (1.91)
ImpX Male 6.62 (1.34)
Female 7.17 (1.67)
ImpY Male 6.75 (1.66)
Female 7.82 (2.36)
ImpC Male 6.75 (1.41)
Female 7.50 (2.17)
Note: Standard deviations are in parentheses.
hard). A significant main effect for task-difficulty was found, F(1.51, 165.95) = 164.264, p < .001, η2 = .599 (Greenhouse-Geiser corrected), indicating that
participants needed increasingly more time to find correct shapes the more difficult this task was (for means and standard deviations see Table 4).
Questionnaire. One of the questions at the end of the experiment asked men and women to rate on a 7-point Likert scale whether they think that members of the other sex perform better or worse than their own sex. A t-test examining sex
differences on the response to this question was highly significant, indicating that both, men and women, believed that females perform worse on this test than males, t(116) = 9.843, p < .001 (men’s expectancy of women’s performance: M = 3.32, SD = 1.19; women’s expectancy of men’s performance: M = 5.37, SD = 1.07). At the end of the study all participants were asked to recall their goal and implementation
intentions by writing them down. These written plans were scored for accuracy by the experimenter on a three-point scale from 0 (not remembered at all) and 2 (verbatim remembered). Most participants remembered their goals and implementation
intentions though often not literally (M = 1.10, SD = .68).
Table 4
Average time taken (in seconds per solution) to select a correct shape for each of the three difficulty levels of men and women in all four goal-conditions during the main block.
Task-difficulty Goal-condition Sex M
Easy Goal Male 5.76 (1.06)
Female 6.49 (1.61)
ImpX Male 6.00 (.99)
Female 6.55 (1.37)
ImpY Male 6.26 (1.32)
Female 7.02 (1.88)
ImpC Male 6.00 (1.22)
Female 6.96 (2.17)
Average Goal Male 6.07 (1.40)
Female 6.56 (1.56)
ImpX Male 6.58 (1.53)
Female 7.11 (1.66)
ImpY Male 6.58 (1.64)
Female 7.51 (2.40)
ImpC Male 6.48 (1.39)
Female 7.37 (2.29)
Hard Goal Male 7.17 (1.70)
Female 8.42 (2.70)
ImpX Male 7.39 (1.66)
Female 8.05 (2.27)
ImpY Male 7.50 (2.27)
Female 9.10 (2.94)
ImpC Male 7.93 (1.90)
Female 8.35 (2.38)
Note: Standard deviations are in parentheses.
Discussion
The data of this study clearly demonstrates that the wording of the then-part of an implementation intention has a significant impact on the effectiveness of the implementation intention. Even though the simple plans to rotate the shapes around either the x- or y-axis were shown to generate an implementation intention effect in form of improved performance, the combination of both plans into a single, albeit complex plan did not help the participants to find more correct shapes. However these findings only hold true for male participants. Female subjects were neither able to profit from the simple implementation intentions nor did they suffer a reduced performance from forming a complex plan. They displayed a similar performance throughout all four goal-conditions. Compared to the male participants, women found fewer correct shapes and took more time to identify these. This is consistent with previous studies that used the mental rotation task and found strong and reliable gender differences in the performance on this task (Cherney & Neff, 2004; Linn &
Petersen, 1985; Vandenberg & Kuse, 1978; Voyer, Voyer, & Bryden, 1995).
The complete absence of an implementation intention effect for female participants was unexpected though. Several studies have shown that men and women use different brain areas to solve mental rotation tasks (Jordan,
Wuestenberg, Heinze, Peters, & Jaencke, 2002; Thomsen et al., 2000), which is rather considered to be evidence for different strategy use between the sexes than for genuine gender differences. Support for this notion comes from studies that manipulated the use of strategy (Glueck, Machat, Jirasko, & Rollett, 2002) or the beliefs about gender differences in spatial tasks (Moè & Pazzaglia, in press) and were thus able to reduce performance differences or even dissolve them completely.
On the other hand there are scientists who favor the explanation of biological factors that are responsible for the inferior performance of women in this kind of tasks, factors such as genetics, brain lateralization or sex hormones (Hampson, 1990;
Kimura, 1999; McGee, 1979; Rilea, Roskos-Ewoldsen, & Boles, 2004). The nature- nurture discussion is far from finished, but judging from the findings of some of the research groups cited here, providing women with a strategy should have had an impact on their performance. But the type of strategy also seems to be important. It was found that men mostly use visual-spatial strategies in which the whole object is being rotated in the mind while women more often report the use of verbal or analytic
strategies in which the object is split into several components which are then rotated individually (C. E. Bethell-Fox & Roger N. Shepard, 1988; E. Pezaris & M. B. Casey, 1991; Schultz, 1991; Thomsen et al., 2000). Dividing the rotation of the object into several steps is of course the more time consuming and more error-prone task. In the study reported here all three implementation intentions specified strategies for visual rotation of the object, because the complexity of the then-part of implementation intentions was the issue of this study and not sex differences in strategy use. The idea was not to close the gap between men’s and women’s performance, but to find out whether the effectiveness of simple and complex implementation intentions is different. Whether men and women respond differently to these types of plans was only an additional object of investigation. The strategies were the same for male and female participants, so that the performance of the subjects could be compared within as well as between the sexes. It was surprising though to find that women did not profit at all from forming implementation intentions. A possible reason for that may be stereotype threat (Steele, 1997). Steele and Aronson describe stereotype threat as “being at risk of confirming, as self-characteristic, a negative stereotype about one's group“ (Steele & Aronson, 1995, p. 797). Stereotype threat has been shown to reduce the performance of members of ethnic groups like African Americans (Davis, Aronson, & Salinas, 2006; Mayer & Hanges, 2003; Steele &
Aronson, 1995) and also of women in stereotypic achievement situations like math tasks (Spencer, Steele, & Quinn, 1999) or spatial rotation tests (Martens, Johns, Greenberg, & Schimel, 2006). Even though in the instructions of this study no special emphasis was put on being an intellectual or academic ability test, it was also not tried to avoid stereotype threat by explicitly stating that the used test would not
produce gender differences, which has been shown to be a reliable method to reduce stereotype threat (Cadinu, Maass, Frigerio, Impagliazzo, & Latinotti, 2003; Martens, Johns, Greenberg, & Schimel, 2006; Spencer, Steele, & Quinn, 1999). The highly significant difference found on the question about performance expectancies of the other sex indicate that men and women were aware of the stereotype that women are supposed to have trouble solving mental rotation tasks. Therefore stereotype threat could very well have been a major influence on the performance of women in this study. Further investigation would be necessary to confirm this assumption, since the data from this study is not sufficient to test this.
Future research on the simple versus complex implementation intention idea should try to avoid tasks with stereotypic performance differences. It suggested to either use a task that is less sensitive to stereotype threat in future studies or at least to instruct all participants that the mental rotation test does not produce gender differences.
Even though the overall effect of implementation intentions was significant for male participants, this study was not able to show that implementation intentions led to a superior performance of the subjects when the rotation axis of the shapes matched the strategy in the implementation intention as was hypothesized. The performance means (Figure 2) indicate that there was an artifact in the random distribution of participants to the different goal-conditions that could not be resolved by an analysis of covariance. Although this artifact did not have a significant impact on the analysis, it can be assumed that it at least weakened the results, as was demonstrated in the repeated measures ANCOVA of the interaction of rotation type and goal-condition. Thus, if the impact of implementation intentions on rotation specific performance was only weak, it might have been reduced to non-significance by this artifact. But aside from that it is important to note that the simple
implementation intentions had a significant effect on the performance of the subjects in this study while the complex implementation intention did not.
For the response time data only the main effects for gender and task difficulty were significant. That is, women were slower than men on the mental rotation task, which has been found by other research groups as well (Bryden, George, & Inch, 1990; Kail, Carter, & Pellegrino, 1979; Tapley & Bryden, 1977), and all participants needed more time to solve the tasks when the angular difference between the shapes and the criterion figure was greater, which is a typical finding in mental rotation studies (Cooper, 1975; Jordan, Wuestenberg, Heinze, Peters, & Jaencke, 2002; Vandenberg & Kuse, 1978). Response time did not interact with the goal condition. Previous studies have shown that individuals who furnish their goal intentions with implementation intentions not only perform the specified behavior more often than those who just have a goal intention, but they also initiate the
specified behavior faster than the those without implementation intentions (Gawrilow, 2004; Gollwitzer, 1999; Sheeran & Silverman, 2003). In addition Hull (1943; 1951) has suggested that the automatization of complex behaviors should be more difficult than the automatization of simple behaviors. Thus, an interaction of response time
and goal condition would have been expected. A reason why this was not found may be the fact that response time is not an independent measure in this study. Since there are two within factors, task-difficulty and rotation-axis, with three levels each, a lot of trials were needed to gather enough response data for each cell of this within factors design. Therefore the participants were told to finish each trial within 15 seconds, so that they wouldn’t get fatigued due to excess duration of the computer experiment. Under these circumstances reaction time is moderated by the time- pressure that was put on the participants in the instructions. Thus the reaction time effects that were found might be even stronger without that time limit and additional effects, like an interaction of goal-condition with reaction time, might have been found.
Conclusion
The goal of this study was to show, that the phrasing of an implementation plan could influence its effectiveness on the performance of the participants. At least for male participants this goal was accomplished. Female participants were not able to profit from forming any kind, simple or complex, implementation intentions. This may be attributed to stereotype threat, since the task used in this study is known to impede the performance of females. Future research on this matter should try to avoid tasks that are stigmatized to favor either sex or reduce possible stereotype threat by using according instructions.
Study 2: Effect of positive and negative incentive on implementation intentions
Overview
The goal of this experiment was to find out whether a low incentive to perform an action can impede the effectiveness of implementation intentions on goal-striving.
According to Heckhausen (1977, p. 175) an action is not carried out just to achieve a goal, but rather to savor or to avoid the anticipated consequences of goal-
achievement, depending on whether these consequences have a positive or negative incentive for the actor. Quality and strength of these incentives are subject to the motives of the actor. Furthermore, Heckhausen (1977, p. 177) proposes that the situation in which a goal-directed action can be performed also offers incentive value.
Meaning that the situation invites a person to control the anticipated course of events in a way that benefits this person’s motives. Combining the personal and situational incentives leads to a resulting tendency to act, which in turn determines the
motivation to act and whether an action is initiated or not. Thus, according to Heckhausen’s model, an experimental manipulation that lowers or raises the incentive should decrease or increase a person’s tendency to act respectively.
This could for example be accomplished by so called punishment, a procedure from instrumental conditioning, which reduces the incentive to perform a specific behavior. The basic punishment procedure involves presenting an aversive stimulus after a specified response. The usual outcome of the procedure is that this response becomes suppressed (Domjan, 1993, p. 283). A great variety of aversive stimuli like electric shocks (Foxx, McMorrow, Bittle, & Bechtel, 1986; Hake & Azrin, 1965), a blast of air (Bayroff, 1940; Maier & Klee, 1943), loud noise (Azrin, 1958; Boyd, 1982;
Charlop, Burgio, Iwata, & Ivancic, 1988), verbal reprimands (Hall et al., 1971), and a squirt of lemon juice in the mouth (Sajwaj, Libet, & Agras, 1974) have been shown to be effective in the punishment of behaviors. Also the research on learned
helplessness (Maier & Seligman, 1976; Seligman, 1975) has demonstrated that exposing humans and animals to inescapable and unavoidable aversive stimuli leads to behavior passivity, the failure or slowness of the organism to initiate actions.
Thus, it is hypothesized, that if an individual is exposed to an unavoidable aversive stimulus after each execution of a goal-directed behavior, this individual will hesitate or maybe even refuse to continue performing this behavior. The response
latency of individuals facing this situation will slow down and more incorrect responses will be made, if incorrect responses are not followed by the aversive stimulus.
Over and above that, presenting an aversive stimulus after a correct performance of the goal-directed behavior should also have an impact on the
strength of the goal-intention. Participants facing punishment for performing the goal- directed behavior should lose interest in reaching the goal and thus show a
diminished intention to achieve this goal. As Sheeran, Webb and Gollwitzer (2005, experiment 1) have recently demonstrated, that the strength of the goal intention moderates the behavioral effects of implementation intentions. In their experiment, Sheeran et al. measured the strength of the goal to study by asking students to indicate how many hours of independent studying they intend to undertake during the following week. Then one half of the students added an implementation intention to their goal intention indicating when and where they intended to study. After the week had passed Sheeran and his colleagues asked the students how many hours they actually did study during the past week. The analysis of this data revealed that the effect of the implementation intention on the amount of hours spent studying was greater the stronger the initial goal-intention had been. That is the effect of the implementation intention on behavior was moderated by the strength of the goal- intention.
For the present study it is therefore assumed that forming an implementation intention will not help to reverse or overcome the behavior passivity that is caused by punishing the goal-directed behavior. Implementation intentions will only help
participants to improve their performance when the incentive to act is high and the strength of the goal-intention is not impaired. So it is expected that only when the incentive is high, participants using implementation intentions will be faster in their responses to the critical stimulus than participants in the other goal conditions. When the incentive is low, no differences in the response latency are expected between the goal-condition groups.
Method
Participants and design. 116 students of University of Konstanz participated in this study in exchange for 2.50 Euros (approximately 3 US$) or course credit (only Psychology students). The distribution of the sexes was unbalanced in favor of
female participants (67 women vs. 49 men). The mean age of the participants was 22 years.
Eight participants were excluded from the data analysis because two of them did not follow the instructions while the remaining six did not remember whether correct or incorrect responses were punished during the experiment. Another three subjects were removed from the analysis since their overall response times were three standard deviations above their groups mean and thus were qualified as outliers. The remaining 105 participants were subjected to a three-factorial design with the two between factors goal-condition (goal, implementation intention vs.
familiarization) and incentive (low vs. high incentive) and the within-factor stimulus type (single-digit critical, single-digit non-critical, multiple-digit ambiguous and multiple-digit non-ambiguous).
Materials. The task of this study was adapted from an experiment by Webb and Sheeran (2004, experiment 3). As in the Webb and Sheeran experiment there were thirteen stimuli consisting of the single-digit critical number 3, four single-digit non-critical numbers (1, 5, 7 and 9), four multiple-digit ambiguous numbers (33, 39, 333 and 413) as well as four multiple-digit non-critical numbers (16, 44, 555 and 694). All stimuli were presented in a black font on a white backdrop on a 15 inch computer screen with a font size of 80 pixels, so that they covered the full foveal field. For the incentive manipulation two sounds, a pleasant and an unpleasant sound from a set of game show sounds were selected. Each sound was trimmed to a length of 500 milliseconds and the volume of both sounds was adjusted to the same level.
The sounds were played using a set of standard desktop stereo speakers.
Procedure. After entering the lab the participants were welcomed by the male experimenter and taken to a cubicle in which the experimental computer was set up.
If two participants appeared at the same time both were taken into the same cubicle and given the introduction together, then one was seated in front of the computer in the first cubicle, while the experimenter took the other participant to the second cubicle in the lab. All instructions were given on a computer screen. The subjects
were allowed to ask questions until after the end of the practice trials, but no longer during the main block. In the instructions the participants were informed that numbers would be presented in the center of the screen one after the other and that they had to indicate how many digits the numbers had by pressing the left or the right mouse button as fast as possible. The response buttons were counterbalanced as was suggested by Webb and Sheeran (2004, p. 416), that is half of the subjects had to press the left mouse button for single-digit stimuli, the other half had to press the right mouse button for this kind of stimuli. Each trial began with the presentation of a
fixation cross in the center of the screen for the duration of 500 milliseconds. After that the fixation cross was replaced by a stimulus that was randomly selected from the stimulus pool. Each stimulus was displayed for a maximum of 1000 milliseconds.
If the participants responded within this time or the time ran out, the number was erased from the screen. Depending on the experimental block either a 1000
millisecond interval followed or one of two sounds was played for 500 milliseconds followed by a 500 milliseconds interval. Each number was presented a total of twelve times, once in each of two practice blocks and ten times in the main block.
In the first practice session the participants were asked to practice pressing the correct mouse buttons. In the second practice session the sounds for correct and incorrect responses were introduced. Accuracy feedback was given to the subjects in both practice sessions during the 1000 milliseconds interval. Accuracy feedback was especially important in the second practice session, because the participants had to learn, which response, correct or incorrect, was followed by what kind of sound, pleasant or unpleasant. During the main block no accuracy feedback was given anymore.
After the second practice session all participants were asked to set
themselves the goal to always press the correct button as fast as possible (‘I always want to press the correct button as fast as possible’). Over and above that one third of the participants were asked to form the implementation intention: ‘If I see the number 3, then I will press the correct mouse button particularly fast’ and another third was asked to familiarize themselves with the number 3. To do so, all
participants first had to look at their goal, implementation intention or familiarization instructions for 15 seconds and silently repeat these instructions to themselves. Then all participants in the former two conditions were asked to write down the goal or the implementation intention on a piece of paper that was put in an envelope next to the
